Length- and Usage-Weighted Indices for Representative Route Extraction from Trajectory Data.

Area of Science:

• Transportation Science
• Data Science
• Geographic Information Systems

Background:

• Conventional metrics for vehicle trajectory analysis often overemphasize short road segments, leading to biased similarity assessments and unstable route grouping.
• This bias distorts the representativeness of extracted paths from large-scale urban trajectory data.

Purpose of the Study:

• To introduce novel weighted indices for representative path extraction from large-scale vehicle trajectory data.
• To mitigate the overstatement of short link representativeness inherent in traditional metrics.

Main Methods:

• Developed weighted indices including link passing ratio, average link usage ratio, and path overlap, all adjusted for link length and usage frequency.
• Applied these indices to a dataset of 18,205 real-world urban trajectories for analysis and comparison with conventional metrics.

Main Results:

• Weighted indices reduced short-link bias by 20-30% and increased the stability of representative path grouping by 15-30% compared to conventional methods.
• Confirmed that weighted indices accurately capture structural characteristics of GPS-based trajectories, reflecting stable link usage and overlap patterns.

Conclusions:

• Refined trajectory similarity metrics at the link level significantly improve the accuracy and stability of representative path extraction.
• These advancements have direct implications for intelligent transportation systems, enhancing applications like vehicle kilometers traveled estimation and congestion diagnostics.